3DL_U-Net model

Model author: Mustapha EL AMMARI

Model description

The model has been designed by transfer tuning then fine tuning 3D U-Net model as autoencoder, using a home made dataset [1] composed of 3D image stack acquired using a confocal microscope. Training and Inference Notebooks are hosted on our Github repo [2].

Stardist Training parameters

• patch size: (64,64,64)
• batch size: 64
• epochs : 200
• image normalization: normalize channel independantly

Training dataset parameters

• split : Train 0.8 / Val 0.2

Inference

• patch size : (64,64,64)
• model : file.h5

References

• [1] Dataset Project: Zenodo
• [4] 3D U-Net Github Project: Github